Thanx to you too
FatChoi
...
Which software would you advice me for hedge fund analysis?
-
Maggette
- Posts: 0
- Joined: Thu Jan 01, 2004 12:00 am
Which software would you advice me for hedge fund analysis?
Ich kam hierher und sah dich und deine Leute lächeln, und sagte mir: Maggette, scheiss auf den small talk, lass lieber deine Fäuste sprechen...
-
astar
- Posts: 0
- Joined: Thu Jan 01, 2004 12:00 am
Which software would you advice me for hedge fund analysis?
Re R vs Octave
I have programmed quite a bit in both R/S+ as well as Octave/Matlab (but quite a while ago) and here are some thoughts. R is really good for statistical programming, because pretty much all recent statisticians, bayesian or otherwise contribute code to the CRAN. Also, if you have to link C libraries, the process is pretty well documented and (fairly) painless. (well, interfacing languages is always crap work)
Octave has better matrix type operations and, if you are using the non-free (as in beer) matlab, it has even better numerical type libraries. So if you need to solve non linear programs, you shd probably use octave. These do crop up from time to time, say in a lasso type estimation. There are very subtle language differences between matlab and octave, though; and any matlab code you might get on the web will need to be debugged in octave.
One irritation in R I found was its habit of resizing matrices. Say I am working with a generic m by n matrix, and the parameter I pass in has n = 1, then (under certain conditions) the system would interpret the matrix as a row or column vector and you matrix multiplications would fail because of dimension mismatch. This arises more often than you would think, particularly if you do some outlier elimination before your regression / L2 estimation step.
I have programmed quite a bit in both R/S+ as well as Octave/Matlab (but quite a while ago) and here are some thoughts. R is really good for statistical programming, because pretty much all recent statisticians, bayesian or otherwise contribute code to the CRAN. Also, if you have to link C libraries, the process is pretty well documented and (fairly) painless. (well, interfacing languages is always crap work)
Octave has better matrix type operations and, if you are using the non-free (as in beer) matlab, it has even better numerical type libraries. So if you need to solve non linear programs, you shd probably use octave. These do crop up from time to time, say in a lasso type estimation. There are very subtle language differences between matlab and octave, though; and any matlab code you might get on the web will need to be debugged in octave.
One irritation in R I found was its habit of resizing matrices. Say I am working with a generic m by n matrix, and the parameter I pass in has n = 1, then (under certain conditions) the system would interpret the matrix as a row or column vector and you matrix multiplications would fail because of dimension mismatch. This arises more often than you would think, particularly if you do some outlier elimination before your regression / L2 estimation step.
-
polysena
- Posts: 0
- Joined: Thu Jan 01, 2004 12:00 am
Which software would you advice me for hedge fund analysis?
I've heard good things about Dalgaard's book. If I recall correctly I only understood all the different *apply (lapply, sapply, tapply, etc.) functions after reading it there! Don't forget that you can browse through books at amazon and google books!
I am not sure I have understood the full of it.. thanks for the cool tip.
I am not sure I have understood the full of it.. thanks for the cool tip.
И ветер, и дождик, и мгла Над холодной пустыней воды.
-
AlphaQuant
- Posts: 1
- Joined: Thu Jan 01, 2004 12:00 am
Which software would you advice me for hedge fund analysis?
astar: Do you know about the drop=FALSE argument to the [], aka slicing, operator? That will stop 1-dimensional result sets being automatically turned into vectors. However I would agree that having to specify this option manually all the time can be quite annoying and prone to error. It would be nice to have a global option where one could change the default behaviour from drop=TRUE to drop=FALSE. (Note that that I haven't used R in a while so if this has changed then so much the better but it didn't use to be available before.)
My 2 cents worth on the whole R vs Matlab vs ... debate follows:
R:
* overall probably my favourite
* data.frames and row and column references by name are such a small feature but really made my life so much easier compared to Matlab.
* It's slicing operations are the most powerful I know (with only one thing missing and that's a built in construct like "end" in Matlab).
* The R package and help system is great
* The community is very active and the packages you can get are just amazing.
* It's free and open source
* It supports functional programming with closures and lack of side-effects (latest buzz phrase but which is actually really handy once you are used to it). Caveat emptor: You need to learn how to use this properly and make use of the vectorized operations or you will take a severe performance hit.
- My biggest disappointment is that the actual code can end up looking quite ugly. I don't really know why because the language features are really powerful but there's just something about the syntax.
Matlab:
* The IDE is very nice and does aid productivity.
* The object oriented parts of the language seem more intuitive to me than those of R (and probably would be for anyone coming from a classical OOP background).
* Nice notation for entering matrices at the command prompt using ";" to separate rows.
- I really really dislike the use of parentheses () for slicing.
- It's expensive, especially once you start using toolboxes.
- Lack of data.frames.
Python:
* A really fantastic language for general programming.
* Easy, intuitive and produces clear and easy to understand code.
* Very nice object orientation.
* Support for array based and numerical programming through Numpy, SciPy, Matplotlib.
* Good support in the scientific community.
- Unfortunately I haven't really managed to get into Numpy and SciPy yet because I find working at the command prompt with multi-dimensional Numpy arrays pretty cumbersome (the nested list notation just produces too many brackets).
- General Python array slicing operations aren't as powerful as in R although I believe there may be some Numpy tools to help with this but I don't know them yet.
My 2 cents worth on the whole R vs Matlab vs ... debate follows:
R:
* overall probably my favourite
* data.frames and row and column references by name are such a small feature but really made my life so much easier compared to Matlab.
* It's slicing operations are the most powerful I know (with only one thing missing and that's a built in construct like "end" in Matlab).
* The R package and help system is great
* The community is very active and the packages you can get are just amazing.
* It's free and open source
* It supports functional programming with closures and lack of side-effects (latest buzz phrase but which is actually really handy once you are used to it). Caveat emptor: You need to learn how to use this properly and make use of the vectorized operations or you will take a severe performance hit.
- My biggest disappointment is that the actual code can end up looking quite ugly. I don't really know why because the language features are really powerful but there's just something about the syntax.
Matlab:
* The IDE is very nice and does aid productivity.
* The object oriented parts of the language seem more intuitive to me than those of R (and probably would be for anyone coming from a classical OOP background).
* Nice notation for entering matrices at the command prompt using ";" to separate rows.
- I really really dislike the use of parentheses () for slicing.
- It's expensive, especially once you start using toolboxes.
- Lack of data.frames.
Python:
* A really fantastic language for general programming.
* Easy, intuitive and produces clear and easy to understand code.
* Very nice object orientation.
* Support for array based and numerical programming through Numpy, SciPy, Matplotlib.
* Good support in the scientific community.
- Unfortunately I haven't really managed to get into Numpy and SciPy yet because I find working at the command prompt with multi-dimensional Numpy arrays pretty cumbersome (the nested list notation just produces too many brackets).
- General Python array slicing operations aren't as powerful as in R although I believe there may be some Numpy tools to help with this but I don't know them yet.
-
Tradenator
- Posts: 0
- Joined: Thu Jan 01, 2004 12:00 am
Which software would you advice me for hedge fund analysis?
[url=/Show%20Post.aspx?PostIDKey=89791]matlab vs R[/url] [url=/Show%20Post.aspx?PostIDKey=86689]Help with R[/url]
I find the zoo package to be pretty useful but there are some limits to it. I found R-intro.pdf to be a good starting point for general R use, there is an example session in an appendix that gives you a good look at all.
I find the zoo package to be pretty useful but there are some limits to it. I found R-intro.pdf to be a good starting point for general R use, there is an example session in an appendix that gives you a good look at all.